On 2025-04-22 05:29, Vagrant Cascadian wrote:
On 2025-04-12, Vagrant Cascadian wrote:
On 2025-04-12, Johannes Schauer Marin Rodrigues wrote:
Luckily, running flash-kernel manually fixed the issue. But had I not
noticed
that /boot/boot.scr still contained a version of a kernel that I had
just
removed, my system would've become unbootable.
Presuming this isn't some bizarre fluke, then this bug is likely
present
in most versions of flash-kernel, as that code has not been touched
for
at least a 2-5 years...
I vaguely recall a bug or merge request coming from Ubuntu that might
be
related...
Does this by chance help at all:
https://salsa.debian.org/installer-team/flash-kernel/-/merge_requests/41
It is about partially installed kernels,
Is it? The LP issue says that it's about "if the kernel and flash-kernel
packages are updated at the same time" or in the MR text "prevent
flash-kernel from "missing" runs when both it and the kernel are
upgraded in a single run of apt." -- that sounds *exactly* like the
issue I observed.
but perhaps might fix this as a
side-effect?
I'm traveling today so I cannot test this right now. Luckily the 3
commits apply cleanly on top of 3.109 so I'll later build myself a
patched flash-kernel and see if this fixes the issue given my reproducer
from earlier.
This is certainly worth looking into at some point ... although maybe a
bit too large changeset during the bookworm freeze right now...
We are again deep into the freeze but it seems that Ubuntu has already
tested these patches in Jammy. We can also give them more testing by
adding them into the MNT repos on top of the patches with which we
already are patching flash-kernel there.
Thanks!
cheers, josch
P.S.: I totally forgot to create a MR now that the dtb of my A311D
Reform has been upstreamed:
https://salsa.debian.org/installer-team/flash-kernel/-/merge_requests/68
